EuroHPC JU Large Scale Access to AI Factories for Industrial AI Innovation with Continuously Open GPU Allocations

This is the EuroHPC access mode built specifically for AI workloads, and it is the most direct route for European researchers who need serious compute without a Horizon Europe grant behind them. Allocations are made in node hours across four petascale and pre-exascale systems - LUMI in Finland, Leonardo in Italy, MareNostrum5 in Spain and MeluXina in Luxembourg - for six-month access periods. The stated scope is explicit about modern AI: ethical artificial intelligence, machine learning, and cutting-edge foundation models and generative AI including large language models, which means training and fine-tuning runs are in scope rather than being squeezed into a traditional simulation-oriented HPC framing. Eligibility is deliberately wide. All scientific users qualify whether or not their work is funded by national or European programmes, as do public sector users, and industrial users participating in EU-funded research and innovation projects such as Horizon Europe or the Digital Europe Programme. The one important boundary: industrial users not involved in an EU research project are directed instead to the AI for Industrial Innovation access modes, so a commercial applicant with no EU project affiliation should check which door they are meant to use before drafting. The call operates on rolling cut-offs rather than a single deadline, with 2026 cut-offs at 27 February, 30 April, 30 June, 31 August, 30 October and 11 December, each at 10:00 CET. That cadence means an unsuccessful application can be revised and resubmitted roughly two months later, and a team that misses one window is never waiting long for the next. As with all compute schemes, no money is awarded - staff and data costs must come from elsewhere.

The EuroHPC JU Large Scale Access call provides industrial users with rapid access to AI Factories' GPU compute resources across European supercomputers (including MareNostrum5, LUMI, Leonardo, JUPITER, and others) for high-impact, high-gain innovative AI research. Cutoffs are rolling and bi-weekly through 2026 and into 2027 (example cutoffs: 30 Jan 2026, 16 Feb 2026, 27 Feb 2026, continuing through 4 Jan 2027). Minimum 50,000 GPU hours per proposal. Industry users requiring large compute for training foundation models, fine-tuning multimodal models, or running large inference experiments can submit on a continuously open basis. Academic and public-sector users should apply via the separate 'AI for Science' access mode. EuroHPC JU is overseeing 19 AI Factories and 13 AI Factory Antennas across Europe to provide free, customized support to SMEs and startups. Strong fit for European industrial AI labs and SMEs needing sovereign EU compute for training and serving large models.

NVIDIA Graduate Fellowship Program is a grant from NVIDIA providing up to $60,000 per award to PhD students conducting research that advances accelerated computing and its applications. Now in its 25th year, the program invites nominations from doctoral students pushing the boundaries of artificial intelligence, robotics, autonomous vehicles, and related fields. Recipients receive not only research funding but also access to NVIDIA technology, products, and engineering expertise, along with a mandatory in-person summer internship. Students are nominated by their faculty advisors and selected based on academic achievement and research area alignment.

Nebius Research Grants Program is sponsored by Nebius. The Nebius research grants program provides researchers with free access to GPU cloud, AI models, and Tavily's agentic search engine, along with Nebius' expertise to accelerate discovery and deployment in AI. It supports academic or scientific projects with clear objectives, innovative use of GPUs for AI, machine learning, data analytics or HPC, and demonstrated feasibility and potential impact.